Artificial intelligence multi-dimensional dynamic weight soil ecological risk whole-process evaluation system

Abstract

This invention relates to the field of environmental risk assessment, specifically to an artificial intelligence-based multidimensional dynamic weighted soil ecological risk assessment system. This system is a full-process soil ecological risk assessment platform, comprising modules for data management, screening value calculation, risk assessment, and risk characterization. The data management module acquires pollutant concentrations and toxicity parameters; the screening value calculation module calculates ecological screening values ​​using either the species sensitivity distribution method or the assessment factor method; the risk assessment module is divided into two stages: the first stage compares concentrations with screening values, and the second stage conducts a detailed assessment using the quotient method, probability method, or evidence weight method when risks exist; finally, the risk characterization module generates a risk distribution map, displaying the risk values ​​of points or areas. This constructs a complete risk assessment process from initial screening to detailed assessment, realizing a full-process assessment of soil ecological risk, and the assessment results can directly support subsequent site management decisions.

Description

Technical Field

[0001] This invention relates to the field of environmental risk assessment, specifically to an artificial intelligence-based multidimensional dynamic weighted soil ecological risk assessment system. Background Technology

[0002] Soil pollution poses a potential threat to the ecological environment and human health. Soil ecological risk assessment is a necessary prerequisite for making decisions on large-scale remediation of contaminated sites and has become an important component of environmental risk management.

[0003] Currently, various soil ecological risk assessment tools have been developed both domestically and internationally, such as the ECOBOX tool from the United States, the CSOIL model from the Netherlands, and the EEC-SSD model developed domestically. However, these tools mainly focus on single aspects such as the derivation of soil ecological screening values, lacking a comprehensive assessment of soil ecological risk throughout the entire process. The assessment results are therefore insufficient to directly support subsequent site management decisions. Furthermore, existing assessment models often utilize foreign toxicity data, failing to fully reflect the protection of native receptors in China, resulting in a lack of local applicability of the assessment results.

[0004] The species sensitivity distribution method, a national ecological and environmental benchmark calculation software developed by the Chinese Research Academy of Environmental Sciences, is one of the more advanced technologies currently available. This software uses four models—normal distribution, log-normal distribution, logistic distribution, and log-logistic distribution—to fit the species sensitivity distribution of toxicity data, providing standardized and regulated technical support for the development of national ecological and environmental benchmarks. However, the software still suffers from limitations such as limited functionality and incomplete assessment processes, failing to meet the comprehensive risk assessment needs of complex sites. Summary of the Invention

[0005] The purpose of this invention is to provide an artificial intelligence-based multidimensional dynamic weighted soil ecological risk assessment system to solve the technical problems existing in the prior art, such as incomplete assessment process, lack of localized data support, and difficulty in supporting decision-making.

[0006] This invention discloses an artificial intelligence-based multidimensional dynamic weighted soil ecological risk assessment system, comprising:

[0007] The data management module is used to acquire site pollutant concentration data and toxicity parameter data;

[0008] The screening value calculation module is connected to the data management module and is used to calculate the soil ecological screening value based on the toxicity parameter data, using the species sensitivity distribution method or the assessment factor method.

[0009] The risk assessment module, connected to the data management module and the screening value calculation module, includes a first-stage assessment unit and a second-stage assessment unit.

[0010] The first-stage assessment unit is used to compare the site pollutant concentration data with the soil ecological screening value to generate a preliminary risk assessment result;

[0011] The second-stage assessment unit is used to conduct a detailed risk assessment using the quotient method, probability method, or evidence weight method when the preliminary risk assessment result indicates the existence of risk, and to generate a detailed risk assessment result.

[0012] The risk characterization module, connected to the risk assessment module, is used to generate a site risk distribution map based on the detailed risk assessment results, wherein the risk distribution map includes point risk values ​​or regional risk values.

[0013] Preferably, the species sensitivity distribution method includes:

[0014] The toxicity data corresponding to the selected pollutant and ecological receptor are selected from the toxicity parameter data;

[0015] The toxicity data are fitted using a normal distribution model, a log-normal distribution model, a logistic distribution model, or a log-logistic distribution model to generate a species sensitivity distribution curve;

[0016] The protection ratio is determined based on the land use type, and the concentration value at (1 - protection ratio) on the species sensitivity distribution curve is taken as the soil ecological screening value.

[0017] Preferably, the land use types include industrial and commercial land, residential land, park green space and mines, with corresponding protection ratios of 40%, 50%, 80% and 80%, respectively.

[0018] Preferably, the evaluation factor method includes:

[0019] The toxicity data corresponding to the selected pollutant and ecological receptor are selected from the toxicity parameter data;

[0020] Assessment factors are determined based on the type and quantity of the toxicity data;

[0021] The soil ecological screening value is calculated by dividing the concentration of pollutants in the soil by the assessment factor.

[0022] Preferably, the determination of the evaluation factors includes:

[0023] The assessment factor is 500 when there is at least one set of acute toxicity data L(E)C50 from plants, invertebrates or insects;

[0024] When only single chronic toxicity data for plants or invertebrates are available, the assessment factor is 100.

[0025] When there are two sets of chronic toxicity data that can represent four species, the assessment factor is 50;

[0026] The assessment factor is 10 when there is chronic toxicity data that can represent at least three trophic levels and seven species.

[0027] Preferably, the quotient method includes the assessment factor method, the species sensitivity distribution method, and the exposure model method. The exposure model method is applicable to receptors that are exposed to soil pollution through foraging without direct contact with the soil, and calculates the soil ecological screening value using the following formula:

[0028] ,

[0029] in, The soil ecological screening value is represented by TRVj, which is the toxicity reference value (mg·kg). -1 ·bw·d -1） , Food intake (kg dry weight of food·kg) -1 Fresh weight d -1 ), The proportion of soil intake in total diet, The proportion of biological intake in total diet. The bioaccumulation coefficient is given.

[0030] Preferably, the probability method includes the safe concentration threshold method, the probability density function overlapping area method, and the probability curve method, wherein the calculation steps of the safe concentration threshold method include:

[0031] Generate cumulative distribution curves of toxicity data and cumulative distribution curves of environmental exposure concentrations;

[0032] Calculate the concentration SSD10 at 10% of the cumulative distribution curve of toxicity data and the concentration ECD90 at 90% of the cumulative distribution curve of environmental exposure concentration;

[0033] Calculate MOS = SSD10 / ECD90. When MOS > 1, it is determined that there is no ecological risk; when MOS ≤ 1, it is determined that there is an ecological risk.

[0034] Preferably, the weighted evidence method includes:

[0035] Construct chemical evidence chains, toxicological evidence chains, and ecological evidence chains;

[0036] Calculate the risk of each sub-indicator within the chain of evidence;

[0037] Calculate the risk quotient of each chain of evidence;

[0038] The risk quotients of each chain of evidence are standardized and assigned weights to calculate the overall risk.

[0039] The formula for calculating the risk quotient of the chemical evidence chain is as follows:

[0040] ,

[0041] in, This indicates the proportion of chemical sub-indicators with a risk level less than 1.3 out of all sub-indicators. This indicates the proportion of chemical sub-indicators with a risk level between 1.3 and 2.6 out of all sub-indicators; This indicates the proportion of chemical sub-indicators with a risk level between 2.6 and 6.5 out of all sub-indicators; This indicates the proportion of chemical sub-indicators with a risk level between 6.5 and 13 out of all sub-indicators. This indicates the proportion of chemical sub-indicators with a risk level greater than 13 out of all sub-indicators.

[0042] The formula for calculating the risk quotient of the toxicological evidence chain is:

[0043] ,

[0044] in, This indicates the proportion of sub-indicators with a toxicological risk of less than 0.7 out of all sub-indicators. This indicates the proportion of toxicological sub-indicators with a risk level between 0.7 and 1 out of all sub-indicators. This indicates the proportion of sub-indicators with a risk level between 1 and 2 out of all sub-indicators. This indicates the proportion of sub-indicators with a toxicological risk level between 2 and 3 out of all sub-indicators. This indicates the proportion of sub-indicators with a risk greater than 3 out of all sub-indicators.

[0045] The formula for calculating the risk quotient of the ecological evidence chain is as follows:

[0046] ,

[0047] in, Ecological indicators of soil pollution risk. The ecological risk indicators of the control soil;

[0048] In order to calculate the overall risk, the risk quotients of each chain of evidence need to be standardized:

[0049] ,

[0050] in, For the standardized HQ of the a-th chain of evidence, As the original risk factor in this chain of evidence, The minimum HQ value for each chain of evidence (100 for chemical evidence chain, 70 for toxicological evidence chain, and 0 for ecological evidence chain). The maximum value of HQ for each chain of evidence (2700 for chemical evidence chain, 800 for toxicological evidence chain, and 1 for ecological evidence chain).

[0051] The formula for calculating the comprehensive risk is as follows:

[0052] ,

[0053] in, Considering the combined risks of the three chains of evidence, Let be the weight of the a-th evidence chain, with the chemical and toxicological evidence chains having a weight of 1, and the ecological evidence chain having a weight of 1.2.

[0054] Preferably, the formula for calculating the risk of the chemical evidence chain sub-indicators is as follows:

[0055] ,

[0056] in, Risks associated with chemical indicators The values ​​are for chemical component indicators. The values ​​of chemical sub-indicators at the control point. The weight of chemical sub-index i;

[0057] The formula for calculating the risk of each indicator in the chain of toxicological evidence is as follows:

[0058] ,

[0059] in, For the risk of toxicological index k, Let k be the value of the toxicological index. The value of the toxicological sub-index k at the control point. , where k is the weight of the toxicological sub-indicator, and 0.2 is the toxicological induction threshold;

[0060] The formula for calculating the risk of the ecological evidence chain sub-indicators is as follows:

[0061] ,

[0062] in, Ecological indicators of soil pollution risk. denoted as the potential species impact ratio of pollutants, and n as the number of pollutants.

[0063] Preferably, the data management module includes a toxicity database, which contains toxicity parameters from domestic and international ecotoxicological data, relevant data released by national government departments, and reliable source data judged by experts. The toxicity database adopts the following data screening principles:

[0064] Chronic toxicity parameters should be given priority, followed by acute toxicity parameters;

[0065] When there are multiple toxicity parameters for the same toxicity endpoint of the same species, the geometric mean shall be taken;

[0066] The formula for calculating the geometric mean is:

[0067] ,

[0068] Where X1, X2, ..., X n There are n toxicity parameter values.

[0069] The beneficial effects of this invention include:

[0070] 1. A complete risk assessment process, from initial screening to detailed evaluation, has been established, realizing the full-process assessment of soil ecological risks. The assessment results can directly support subsequent site management decisions.

[0071] 2. It integrates multiple assessment methods such as species sensitivity distribution method, assessment factor method, quotient method, probability method and evidence weight method to meet the assessment needs of different types of sites and improve the applicability and accuracy of the assessment;

[0072] 3. It has a built-in comprehensive database containing domestic and international toxicity data and has established a scientific data screening mechanism, which solves the problem of insufficient localization of assessment results;

[0073] 4. By adopting differentiated protection standards based on land use types, refined risk management for different land use types has been achieved;

[0074] 5. An innovative comprehensive risk assessment method for compound pollution has been implemented, improving the ability to assess complex pollution situations at actual sites. Attached Figure Description

[0075] Figure 1 This is the overall architecture diagram of the artificial intelligence multidimensional dynamic weighted soil ecological risk full-process assessment system of the present invention;

[0076] Figure 2 This is a flowchart illustrating the implementation of the species sensitivity distribution method of the present invention.

[0077] Figure 3 This is a flowchart illustrating the implementation of the evaluation factor method of this invention.

[0078] Figure 4 This is a flowchart illustrating the implementation of the quotient method of the present invention;

[0079] Figure 5 This is a flowchart illustrating the implementation of the probability method of this invention.

[0080] Figure 6 This is a flowchart illustrating the implementation of the evidence weighting method of this invention.

[0081] Figure 7 This is an example of a risk distribution map of an industrial site generated by the present invention. Detailed Implementation

[0082] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0083] like Figure 1 As shown, the artificial intelligence multidimensional dynamic weighted soil ecological risk full-process assessment system of the present invention includes a data management module 1, a screening value calculation module 2, a risk assessment module 3, and a risk characterization module 4.

[0084] Data management module 1 is used to acquire site pollutant concentration data and toxicity parameter data. Screening value calculation module 2, connected to data management module 1, is used to calculate soil ecological screening values ​​based on toxicity parameter data, using either the species sensitivity distribution method or the assessment factor method. Risk assessment module 3, connected to data management module 1 and screening value calculation module 2, includes a first-stage assessment unit 31 and a second-stage assessment unit 32. The first-stage assessment unit 31 compares site pollutant concentration data with soil ecological screening values ​​to generate preliminary risk assessment results. The second-stage assessment unit 32, when the preliminary risk assessment results indicate the presence of risk, conducts a detailed risk assessment using the quotient method, probability method, or weighted evidence method, generating detailed risk assessment results. Risk characterization module 4, connected to risk assessment module 3, is used to generate a site risk distribution map based on the detailed risk assessment results, where the risk distribution map includes point risk values ​​or regional risk values.

[0085] In one embodiment of the present invention, the data management module 1 includes a site information input unit, a pollutant concentration data input unit, a toxicity database, and a data preprocessing unit. The toxicity database stores toxicity parameters from domestic and international ecotoxicological data, data released by national government departments, and data from reliable sources determined by experts. These data sources include international databases such as ECOTOX, ETOX, ECHA, and eCHEMPORTAL, as well as research results from domestic research institutions such as the Chinese Research Academy of Environmental Sciences and the Chinese Academy of Sciences.

[0086] Data management module 1 employs scientific data screening principles: prioritizing chronic toxicity parameters, followed by acute toxicity parameters; when multiple toxicity parameters exist for the same toxicity endpoint for the same species, the geometric mean is used. The formula for calculating the geometric mean is:

[0087] ,

[0088] in, There are n toxicity parameter values. This approach effectively avoids the influence of extreme values ​​on the assessment results, improving the stability and reliability of the assessment. For example, when there are multiple NOEC values ​​for lead on earthworms (100 mg / kg, 130 mg / kg, 150 mg / kg), the system will calculate the geometric mean. As the final parameter value.

[0089] The screening value calculation module 2 implements two methods for deriving soil ecological screening values: the species sensitivity distribution method and the evaluation factor method.

[0090] like Figure 2 As shown, the implementation process of the species sensitivity distribution method includes the following steps: First, toxicity data corresponding to the selected pollutants and ecological receptors are screened from the toxicity parameter data; then, the toxicity data are fitted using a normal distribution model, a log-normal distribution model, a logistic distribution model, or a log-logistic distribution model to generate a species sensitivity distribution curve; finally, the protection ratio is determined according to the land use type, and the concentration value at (1 - protection ratio) on the species sensitivity distribution curve is taken as the soil ecological screening value.

[0091] In a preferred embodiment of the invention, the system supports fitting multiple statistical models and automatically selects the optimal model by calculating the goodness of fit (R²). For example, when assessing chromium pollution, the R² value of the log-normal distribution model is 0.96, which is higher than other models. Therefore, the system will select the log-normal distribution model to generate the species sensitivity distribution curve.

[0092] Land use types include industrial and commercial land, residential land, park green space, and mining, with corresponding protection ratios of 40%, 50%, 80%, and 80%, respectively. This differentiated protection standard fully considers the actual needs of different land use types. For example, industrial and commercial land typically has less human activity and relatively lower ecological protection requirements, so its protection ratio is set at 40%; while park green space, as an important component of the urban ecosystem, has higher protection requirements, and its protection ratio is set at 80%.

[0093] Taking a copper-contaminated site as an example, assuming the site is planned to be used as a park green space with a protection ratio of 80%, the HC20 (i.e. the hazard concentration that affects 20% of species) calculated by the system through the species sensitivity distribution method is 72 mg / kg. This value is the soil ecological screening value for copper contamination at the site.

[0094] like Figure 3 As shown, the implementation process of the assessment factor method includes the following steps: First, toxicity data corresponding to the selected pollutants and ecological receptors are screened from the toxicity parameter data; then, assessment factors are determined according to the type and quantity of toxicity data; finally, the soil ecological screening value is calculated by dividing the concentration of pollutants in the soil by the assessment factors.

[0095] The assessment factor is determined based on the sufficiency and uncertainty of toxicity data: the assessment factor is 500 when there is at least one set of acute toxicity data (L(E)C50) from plants, invertebrates, or insects; 100 when there is only a single set of chronic toxicity data from plants or invertebrates; 50 when there are two sets of chronic toxicity data representing four species; and 10 when there are at least three trophic levels and seven species. A higher assessment factor value indicates higher uncertainty and a greater safety factor.

[0096] For example, for a nickel-contaminated site, if there is only one set of LC50 data for plants (250 mg / kg) and the evaluation factor is set to 500, then the soil ecological screening value for nickel at this site is 250 / 500 = 0.5 mg / kg.

[0097] Risk assessment module 3 includes a first-stage assessment unit 31 and a second-stage assessment unit 32.

[0098] The first-stage assessment unit 31 conducts preliminary risk screening by comparing the site pollutant concentration data with the soil ecological screening values ​​calculated by the screening value calculation module 2. When the pollutant concentration is lower than the screening value, the risk is considered acceptable; when the pollutant concentration is higher than the screening value, it indicates that there may be unacceptable risk, requiring a detailed assessment in the second stage.

[0099] The second-stage assessment unit 32 provides three detailed risk assessment methods: the quotient method, the probability method, and the weighted evidence method.

[0100] The second-stage assessment unit 32 can further implement a multi-dimensional dynamic weight assessment method to improve the assessment accuracy of complex contaminated sites.

[0101] The multidimensional dynamic weighting assessment method is an innovative assessment approach based on matrix theory and topological principles, suitable for sites with multiple coexisting pollutants and complex environmental conditions. This method can dynamically adjust the risk weights of different pollutants, more accurately characterizing the ecological risk status of complex sites.

[0102] Practice has shown that traditional risk assessment methods often treat pollutants as independent entities, failing to fully consider the interactions between pollutants and the impact of environmental conditions on pollutant migration and transformation, leading to discrepancies between assessment results and actual conditions. For example, in the risk assessment of a site with complex heavy metal contamination, the traditional method assessed it as "relatively risky," while actual monitoring data showed a significant reduction in plant and soil biodiversity at the site, indicating that the actual risk level was underestimated.

[0103] The multi-dimensional dynamic weighting assessment method overcomes the aforementioned limitations by constructing a multi-dimensional feature space of pollutants, receptors, and the environment. The specific implementation steps are as follows:

[0104] The first step is to construct a multidimensional feature matrix of pollutants, receptors, and the environment:

[0105] ,

[0106] in: It is a multidimensional feature matrix. For the first The pollutant / receptor and the first Interactive characteristic values ​​of environmental factors (dimensionless, typically between 0 and 1) The total number of pollutants and receptors. This represents the total number of environmental factors. In practical applications, interaction characteristic values ​​can be determined through experimental data or expert scoring. Taking the interaction between copper and pH as an example, The values ​​can be obtained by normalizing experimental data on the bioavailability of copper under different pH conditions. At pH=7.0, the bioavailability of copper is moderate. The value is typically around 0.65; however, under acidic conditions at pH 5.0, the bioavailability of copper increases. It may rise to 0.85. The second step is to perform singular value decomposition on the characteristic matrix:

[0107] ,

[0108] in: for An orthogonal matrix representing the eigenvector space of pollutants and receptors; for The elements on the diagonal of the diagonal matrix are singular values, representing the importance of each feature; for The transpose of an orthogonal matrix represents the eigenvector space of environmental factors. Singular value decomposition (SVD) is a dimensionality reduction technique that extracts the main features of data and reduces computational complexity. Based on practical experience, it is usually sufficient to retain principal components that explain more than 85% of the total variance; this preserves key information while effectively reducing computational load.

[0109] The formula for calculating the static weight vector is:

[0110]

[0111] in: It is a static weight vector. Singular value matrix diagonal front A vector consisting of non-zero singular values For matrix rank, For the first The third step is to construct a topological connectivity matrix to characterize the interactions between pollutants:

[0112] ,

[0113] Where: T is the topological connectivity matrix. pollutants With pollutants The interaction coefficient (dimensionless). When the two pollutants have a synergistic enhancing effect. When antagonistic effects exist When there is no interaction The construction of the topological connectivity matrix is ​​based on extensive toxicological experimental data. For example, there is an antagonistic effect between copper and zinc, and according to existing research, their interaction coefficient... The value typically ranges from 0.85 to 0.95; however, lead and cadmium exhibit a synergistic enhancing effect, with their interaction coefficient... The values ​​are typically between 1.15 and 1.25. In practical applications, these interaction coefficients can be fine-tuned according to the pollutant concentration level to more accurately reflect the pollutant interaction effects under specific site conditions.

[0114] Step 4: Determine the environmental impact vector:

[0115] ,

[0116] middle: This is the environmental impact vector. For environmental conditions on pollutants The influence coefficient of migration and transformation (dimensionless). When environmental conditions promote the migration and transformation of pollutants... When environmental conditions inhibit the migration and transformation of pollutants, When environmental conditions have no significant impact on the migration and transformation of pollutants, The environmental impact coefficient is related to various factors, such as pH value, organic matter content, and redox potential. Taking the effect of pH value on heavy metal migration as an example, when the soil pH value is below 5.5, the activity of most heavy metals increases. The value may be between 1.10 and 1.30; when the pH value is between 6.5 and 7.5, the heavy metal activity is moderate. The value is approximately 1.0; when the pH value is higher than 8.0, the activity of most heavy metals decreases. The value may be between 0.70 and 0.90.

[0117] Step 5: Calculate the dynamic weight vector:

[0118] ,

[0119] in: For dynamic weight vectors, It is a static weight vector. This represents the Hadamard product (element-wise multiplication). This represents the product of the topological connectivity matrix and the environmental impact vector. The Hadamard product operation can capture the interaction between pollutant characteristics and environmental conditions, resulting in a more accurate weight allocation. The dynamic weight vector requires normalization.

[0120] ,

[0121] in: For a normalized dynamic weight vector, Dynamic weight vector The first in Each element.

[0122] Step 6: Calculate the bioavailability correction factor:

[0123]

[0124] in: pollutants Bioavailability correction factor (dimensionless, value range 0-1) This is the rate constant (usually taken between 0.5 and 2.0, determined based on the pH sensitivity of the pollutant). Soil pH value, pollutants pH value at which bioavailability is maximized. Pollutant-specific parameters (in %) -1 (Reflects the ability of pollutants to bind with organic matter) Soil organic carbon content (in %). The bioavailability correction factor considers the effects of pH and organic matter content on pollutant bioavailability. Different pollutants... and The values ​​vary. For example, for copper, It is usually between 5.0 and 5.5. Approximately 0.18-0.22%-1; for cadmium, It is usually between 6.0 and 6.5. Approximately 0.10-0.15% -1 These parameters can be determined experimentally or obtained from existing literature.

[0125] Step 7: Calculate the overall risk index:

[0126]

[0127] in: The comprehensive risk index (dimensionless) Normalized pollutants Dynamic weights, pollutants Risk quotient (dimensionless) pollutants Bioavailability correction factor (dimensionless).

[0128] The comprehensive risk index calculation takes into account the risk quotient of pollutants, dynamic weights, and bioavailability correction, thus providing a more comprehensive reflection of the site's actual risk status. Risk Quotient The calculation follows the same method as the aforementioned quotient method.

[0129] Step 8, Risk Level Determination: Based on the calculated comprehensive risk index The risk level should be determined according to the following criteria:

[0130] Negligible risk;

[0131] Minor risk;

[0132] Less risk;

[0133] Significant risk;

[0134] Serious risks;

[0135] These thresholds were determined based on statistical analysis of a large amount of actual site assessment data. For example, a retrospective assessment of 100 contaminated sites with known risk levels revealed that... Sites with values ​​less than 0.1 have virtually no observable ecological impact; while Sites with values ​​greater than 0.8 generally exhibit significant degradation of ecosystem functions. Therefore, these thresholds have strong practical guiding significance.

[0136] The multi-dimensional dynamic weighting assessment method complements existing assessment methods, jointly constructing a more comprehensive risk assessment system. In practical applications, different assessment methods each have their advantages: the quotient method is simple and intuitive to operate; the probability method can characterize the uncertainty of risk; the evidence weighting method can integrate multiple evidence chains; and the multi-dimensional dynamic weighting assessment method is particularly suitable for sites with multiple pollutants coexisting and complex environmental conditions.

[0137] During the assessment process, appropriate assessment methods can be selected based on the characteristics of the site. For example, for sites with a single pollutant and limited data, the quotient method can be given priority; for sites with abundant data but high uncertainty, the probability method can be given priority; for sites with multiple sources of evidence, the weighted evidence method can be given priority; and for sites with multiple pollutants coexisting and complex environmental conditions, the multi-dimensional dynamic weighted assessment method can be given priority.

[0138] The synergistic application of the multi-dimensional dynamic weight evaluation method with existing methods includes:

[0139] When used in conjunction with the quotient method, dynamic weights can be used to perform a weighted average of the risk quotients of each pollutant, resulting in a more accurate comprehensive risk assessment. For example, in an industrial site assessment, the traditional quotient method assigns risk quotients of 1.2, 1.8, and 0.9 to copper, lead, and zinc, respectively, with a simple average yielding a comprehensive risk of 1.3. However, considering the antagonistic effects of copper and zinc, the multi-dimensional dynamic weight assessment method yields a comprehensive risk of 1.1, which better reflects the actual observed level of ecological impact.

[0140] When used in conjunction with probabilistic methods, dynamic weights can be introduced into the probability distribution model to optimize the calculation of risk probabilities. For example, in the safe concentration threshold method, when calculating the comprehensive MOS value of multiple pollutants, dynamic weights can be used instead of simple averaging. An assessment of a contaminated farmland site showed that the MOS value calculated by the traditional method was 1.2, indicating no risk; however, after considering the synergistic effects of pollutants, the MOS value calculated by the multi-dimensional dynamic weight assessment method was 0.85, correctly identifying the potential risk of the site, which was consistent with subsequent monitoring results.

[0141] When used in conjunction with the weighted evidence method, it can improve the weight allocation of each chain of evidence and enhance the accuracy of the comprehensive assessment. For example, when calculating the comprehensive risk of chemical, toxicological, and ecological evidence chains, traditional methods use fixed weights; while the multi-dimensional dynamic weighting assessment method dynamically adjusts the weights based on the quality and reliability of the evidence, making the assessment results more objective.

[0142] Taking a former chemical plant site as an example, the site contains four heavy metals: copper, lead, zinc, and cadmium. The soil pH is 6.8, and the total organic carbon (TOC) content is 2.5%. A multi-dimensional dynamic weighting assessment method is used for evaluation.

[0143] First, based on site monitoring data and experimental analysis, a multidimensional feature matrix of pollutants, receptors, and the environment is constructed:

[0144] ,

[0145] Singular value decomposition is performed on the characteristic matrix to obtain singular values. Calculate the static weight vector. .

[0146] Considering the antagonistic effect between copper and zinc, and the synergistic enhancing effect between lead and cadmium, a topological connectivity matrix was constructed based on toxicological experimental data:

[0147] ,

[0148] Based on the site's environmental conditions, such as pH value and organic matter content, determine the environmental impact vector. Here, the environmental impact coefficients for lead and cadmium are relatively high, reflecting the relatively high activity of these two elements at pH=6.8.

[0149] Calculate the dynamic weight vector And after normalization, we get .

[0150] Bioavailability correction parameters were determined based on pollutant characteristics: copper. lead Zinc Cadmium The bioavailability correction factor was calculated. .

[0151] The concentrations of various pollutants at the site were: copper 90 mg / kg, lead 220 mg / kg, zinc 150 mg / kg, and cadmium 4.5 mg / kg. The corresponding soil ecological screening values ​​were: copper 72 mg / kg, lead 95 mg / kg, zinc 180 mg / kg, and cadmium 2.6 mg / kg. The calculated risk quotients were... .

[0152] The final calculated comprehensive risk index is as follows: .

[0153] According to the risk level assessment criteria If the value falls within the range of 0.6-0.8, the ecological risk of this site is classified as "relatively high," and risk management measures are required.

[0154] To verify the accuracy of the assessment results, an ecological survey was conducted on the site. The survey found that soil microbial activity was reduced by about 35% compared to normal levels, earthworm density was reduced by about 60% compared to the control area, and plant growth was inhibited by about 40-50%. These indicators all suggest that the site does indeed have significant ecological risks, which is consistent with the assessment results.

[0155] Compared to traditional assessment methods, the multi-dimensional dynamic weighting assessment method demonstrates a clear advantage in this case. The traditional quotient method calculates an average risk quotient of 1.53, failing to reflect the interactions between pollutants; the evidence weighting method assesses "lower risk," underestimating the actual risk level. In contrast, the multi-dimensional dynamic weighting assessment method, by considering the interactions between pollutants and the influence of environmental conditions, yields a more accurate and reliable assessment result.

[0156] Practice has shown that the multi-dimensional dynamic weighting evaluation method is particularly suitable for the following scenarios:

[0157] (1) Complex contaminated sites where multiple pollutants coexist;

[0158] (2) Sites where there are obvious interactions between pollutants;

[0159] (3) Sites where environmental conditions have a significant impact on the migration and transformation of pollutants;

[0160] (4) Sites that require precise assessment to guide remediation decisions.

[0161] In summary, the multi-dimensional dynamic weighting assessment method, a key innovation of this invention, achieves more accurate assessments of complex contaminated sites by introducing matrix theory and topological principles. This method complements existing quotient methods, probability methods, and evidence weighting methods, collectively forming a comprehensive and scientific soil ecological risk assessment system, providing strong support for risk management and decision-making regarding contaminated sites.

[0162] like Figure 4 As shown, the quotient method includes the assessment factor method, the species sensitivity distribution method, and the exposure model method. In the second-stage assessment, these methods were enhanced based on the first-stage method, allowing users to supplement more detailed ecological receptor information and toxicity data, thereby obtaining more accurate soil ecological screening values.

[0163] In particular, the exposure model method is applicable to receptors (such as wild animals or birds) that are exposed to soil pollution through foraging without direct contact with the soil. The soil ecological screening value is calculated using the following formula:

[0164] ,

[0165] in, The soil ecological screening value is represented by TRVj, which is the toxicity reference value (mg·kg). -1 ·bw·d -1） , Food intake (kg dry weight of food·kg) -1 Fresh weight d -1 ), The proportion of soil intake in total diet, The proportion of biological intake in total diet. The bioaccumulation coefficient is given.

[0166] In practical applications, these parameters are typically determined based on literature recommendations. For example, when assessing lead pollution in birds, The possible value is 1.63 mg·kg -1 ·bw·d -1 , The value is 0.02. The value is 0.98. The value is 0.45. Substituting this into the formula, the eco-value of lead is calculated. It is approximately 18.3 mg / kg.

[0167] The formula for calculating risk quotient is:

[0168] Risk Quotient ,

[0169] When the risk quotient is greater than 1, it indicates that there is an unacceptable ecological risk; when the risk quotient is less than or equal to 1, the risk is acceptable.

[0170] like Figure 5 As shown, the probability method includes the safe concentration threshold method, the probability density function overlapping area method, and the probability curve method.

[0171] The calculation steps of the safe concentration threshold method include: generating a cumulative distribution curve of toxicity data and a cumulative distribution curve of environmental exposure concentration; calculating the concentration SSD10 at 10% of the cumulative distribution curve of toxicity data and the concentration ECD90 at 90% of the cumulative distribution curve of environmental exposure concentration; calculating MOS = SSD10 / ECD90, and determining that there is no ecological risk when MOS > 1, and that there is an ecological risk when MOS ≤ 1.

[0172] In one embodiment of the present invention, when assessing arsenic pollution at an industrial site, the SSD10 was obtained as 7.54 mg / kg by fitting toxicity data, and the ECD90 was obtained as 28.0 mg / kg by analyzing the site's arsenic concentration data. The calculated MOS was 7.54 / 28.0≈0.27<1, indicating that the site has a high ecological risk of arsenic pollution.

[0173] The probability density function overlap area method places the probability distribution curves of toxicity data and environmental exposure concentration probability distribution curves on the same coordinate system and calculates the area of ​​overlap between the two curves as the probability of adverse effects on the organism. Specifically, multiple points are taken at equal intervals within the concentration range, and the probability density function values ​​of the two distributions are calculated for each point. The smaller value is then taken, and the overlap area is obtained by integration.

[0174] The probability curve distribution method plots a curve with the cumulative probability of toxicity data on the horizontal axis and the probability of pollutant concentration exceeding the corresponding effect on the vertical axis. The area under the curve represents the potential ecological risk of the pollutant. The larger the area under the curve, the higher the risk.

[0175] like Figure 6 As shown, the weighted evidence method includes constructing chemical evidence chains, toxicological evidence chains, and ecological evidence chains; calculating the risk of sub-indicators within each evidence chain; calculating the risk quotient of each evidence chain; standardizing the risk quotient of each evidence chain and assigning weights to it; and calculating the overall risk.

[0176] The formula for calculating the risk of each indicator in the chemical evidence chain is as follows:

[0177] ,

[0178] in, Risks associated with chemical indicators The values ​​are for chemical component indicators. The values ​​of chemical sub-indicators at the control point. This represents the weight of chemical component index i. Weights are typically determined based on the pollutant's toxicity and environmental behavior characteristics. For example, for heavy metals, chromium might have a weight of 1.2, and cadmium might have a weight of 1.5, to reflect their higher ecotoxicity.

[0179] The formula for calculating the risk of each indicator in the chain of toxicological evidence is as follows:

[0180] ,

[0181] in, For the risk of toxicological index k, Let k be the value of the toxicological index. The value of the toxicological sub-index k at the control point. , where k is the weight of the toxicological sub-indicator, and 0.2 is the induction threshold of the toxicological indicator, meaning that an induction of more than 20% in the response of the toxicological indicator is considered a significant effect. Toxicological indicators typically include biomarkers (such as metallothionein MT), DNA damage indicators (such as 8-hydroxydeoxyguanosine 8-OHDG), etc.

[0182] The formula for calculating the risk of the ecological evidence chain sub-indicators is as follows:

[0183] ,

[0184] in, Ecological indicators of soil pollution risk. denoted by , where is the potential species impact ratio of the pollutant, and 'n' is the quantity of the pollutant. This formula reflects the combined toxic effects of multiple pollutants, rather than a simple additive effect.

[0185] The formula for calculating the risk quotient of a chemical evidence chain is:

[0186] ,

[0187] in, This indicates the proportion of chemical sub-indicators with a risk level less than 1.3 out of all sub-indicators. This indicates the proportion of chemical sub-indicators with a risk level between 1.3 and 2.6 out of all sub-indicators; This indicates the proportion of chemical sub-indicators with a risk level between 2.6 and 6.5 out of all sub-indicators; This indicates the proportion of chemical sub-indicators with a risk level between 6.5 and 13 out of all sub-indicators. This indicates the proportion of chemical sub-indicators with a risk level greater than 13 out of all sub-indicators.

[0188] The formula for calculating the risk quotient of the toxicological evidence chain is:

[0189] ,

[0190] in, This indicates the proportion of sub-indicators with a toxicological risk of less than 0.7 out of all sub-indicators. This indicates the proportion of toxicological sub-indicators with a risk level between 0.7 and 1 out of all sub-indicators. This indicates the proportion of sub-indicators with a risk level between 1 and 2 out of all sub-indicators. This indicates the proportion of sub-indicators with a toxicological risk level between 2 and 3 out of all sub-indicators. This indicates the proportion of sub-indicators with a risk greater than 3 out of all sub-indicators.

[0191] The formula for calculating the risk quotient of the ecological evidence chain is:

[0192] ,

[0193] in, Ecological indicators of soil pollution risk. To assess the ecological risk of the control soil, the risk quotients for each chain of evidence need to be standardized in order to calculate the overall risk.

[0194] ,

[0195] in, For the first Standardized HQ for a chain of evidence The minimum HQ value for each chain of evidence (100 for chemical evidence chain, 70 for toxicological evidence chain, and 0 for ecological evidence chain). The maximum value of HQ for each chain of evidence (2700 for chemical evidence chain, 800 for toxicological evidence chain, and 1 for ecological evidence chain).

[0196] Finally, the comprehensive risk calculation formula is as follows:

[0197] ,

[0198] In one specific embodiment of the present invention, the soil surrounding a metal smelter is assessed to obtain a chemical evidence chain risk quotient. =4500 (Severity Level), Risk Quotient in Toxicological Evidence Chain =567.5 (Severity Level), Ecological Evidence Chain Risk Score =0.97 (higher risk level). After standardization, they are 0.55, 0.68 and 0.97 respectively, with an overall risk EnvRI of 0.75, which is classified as a severe risk level.

[0199] Risk characterization module 4 is responsible for visualizing the assessment results and generating a site risk distribution map. For example... Figure 7 As shown, depending on the assessment method, risk distribution maps include two types:

[0200] 1. Location Risk Value Map: Applicable to the assessment results of the quotient method and the weight of evidence method, the risk level of each sampling point is marked on the map with different colors;

[0201] 2. Regional Risk Value Map: Applicable to probabilistic assessment results, it generates a risk distribution map of the entire site through interpolation, using different color gradients to represent the spatial variation of risk levels.

[0202] Risk levels are typically categorized into five levels: negligible, minor, relatively minor, relatively major, and severe, represented by green, blue, yellow, orange, and red, respectively.

[0203] The system of this invention has been applied in risk assessment of multiple contaminated sites. The following uses the soil around a metal smelter as an example to illustrate the system's workflow and assessment results.

[0204] The soil at this site contained severe levels of heavy metals such as arsenic and lead, and a total of 11 soil samples were collected (including 5 control area samples). Users input site information and pollutant concentration data through data management module 1, and select industrial and commercial land as the future land use type.

[0205] Table 1. Soil pollutant concentration data at the site

[0206]

[0207] Referring to Table 2, the first-stage assessment adopted the species sensitivity distribution method. The system automatically screened relevant toxicity data of arsenic and lead from the toxicity database and used a normal distribution model to fit the data (R² were 0.97 and 0.90, respectively) to determine the soil ecological screening values ​​as follows: arsenic 21.5 mg / kg and lead 284.9 mg / kg.

[0208] According to Table 1, the system compared the site pollutant concentrations with the screening values ​​and found that the arsenic (35.8 mg / kg) and lead (629.05 mg / kg) at point S1 both exceeded the screening values, with risk quotients of 1.67 and 2.21, respectively, indicating that there is an unacceptable risk and a second-stage assessment is required.

[0209] Table 2. Chronic toxicity data of susceptible species

[0210]

[0211] The second phase of the assessment used the weighted evidence method to construct chemical, toxicological, and ecological evidence chains. Calculations showed that the chemical evidence chain risk quotient for site S1 was 4500 (severe level), the toxicological evidence chain risk quotient was 567.50 (severe level), and the ecological evidence chain risk quotient was 0.97 (relatively high level). The overall risk EnvRI was 0.75, classifying it as a severe risk level.

[0212] Ultimately, the system generates a risk distribution map, which visually displays the risk level of each location on the site, providing a scientific basis for subsequent risk management decisions.

[0213] As can be seen from the above examples, the system of the present invention realizes the whole process of risk assessment from initial screening to detailed evaluation, can meet the assessment needs of different types of sites, and the assessment results are intuitive and clear, with good practicality.

[0214] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An AI-powered, multi-dimensional, dynamic weighted, full-process assessment system for soil ecological risk, characterized in that: include: The data management module is used to acquire site pollutant concentration data and toxicity parameter data; The screening value calculation module is connected to the data management module and is used to calculate the soil ecological screening value based on the toxicity parameter data, using the species sensitivity distribution method or the assessment factor method. The risk assessment module, connected to the data management module and the screening value calculation module, includes a first-stage assessment unit and a second-stage assessment unit. The first-stage assessment unit is used to compare the site pollutant concentration data with the soil ecological screening value to generate a preliminary risk assessment result; The second-stage assessment unit is used to conduct a detailed risk assessment using a multi-dimensional dynamic weight assessment method when the preliminary risk assessment results indicate the existence of risk, and to generate a detailed risk assessment result. The multi-dimensional dynamic weight evaluation method includes: The first step is to construct a multidimensional feature matrix of pollutants, receptors, and the environment: ； in: It is a multidimensional feature matrix. For the first The pollutant / receptor and the first Interactive characteristic values ​​of environmental factors The total number of pollutants and receptors. This represents the total number of environmental factors. The second step is to perform singular value decomposition on the characteristic matrix: ； in: for An orthogonal matrix representing the eigenvector space of pollutants and receptors; for A diagonal matrix, where the elements on the diagonal are singular values, representing the importance of each feature; for The transpose of an orthogonal matrix represents the eigenvector space of environmental factors; The formula for calculating the static weight vector is: ； in: It is a static weight vector. Singular value matrix diagonal front A vector consisting of non-zero singular values For matrix rank, For the first One singular value; The third step is to construct a topological connectivity matrix to characterize the interactions between pollutants: ； in: It is the topological connectivity matrix. pollutants With pollutants The interaction coefficient, when the two pollutants have a synergistic enhancing effect. When antagonistic effects exist, When there is no interaction, ; Step 4: Determine the environmental impact vector: ； in: This is the environmental impact vector. For environmental conditions on pollutants The influence coefficient of migration and transformation; when environmental conditions promote the migration and transformation of pollutants. When environmental conditions inhibit the migration and transformation of pollutants, When environmental conditions have no significant impact on the migration and transformation of pollutants, ; Step 5: Calculate the dynamic weight vector: ； in: For dynamic weight vectors, It is a static weight vector. Represents the Hadamard product. This represents the product of the topological connectivity matrix and the environmental influence vector; Dynamic weight vector normalization processing: ； in: For a normalized dynamic weight vector, For dynamic weight vectors The first in One element; Step 6: Calculate the bioavailability correction factor: ； in: pollutants Bioavailability correction factor The rate constant is Soil pH value, pollutants pH value at which bioavailability is maximized. For pollutant-specific parameters, Soil organic carbon content; Step 7: Calculate the overall risk index: ； in: As a comprehensive risk index, Normalized pollutants Dynamic weights, pollutants Risk factor pollutants Bioavailability correction factor; The risk characterization module, connected to the risk assessment module, is used to generate a site risk distribution map based on the detailed risk assessment results, wherein the risk distribution map includes point risk values ​​or regional risk values.

2. The system according to claim 1, characterized in that, The species sensitivity distribution method includes: The toxicity data corresponding to the selected pollutant and ecological receptor are selected from the toxicity parameter data; The toxicity data are fitted using a normal distribution model, a log-normal distribution model, a logistic distribution model, or a log-logistic distribution model to generate a species sensitivity distribution curve; The protection ratio is determined based on the land use type, and the concentration value at (1 - protection ratio) on the species sensitivity distribution curve is taken as the soil ecological screening value.

3. The system according to claim 2, characterized in that, The land use types include industrial and commercial land, residential land, park green space and mines, with corresponding protection ratios of 40%, 50%, 80% and 80%, respectively.

4. The system according to claim 1, characterized in that, The evaluation factor method includes: The toxicity data corresponding to the selected pollutant and ecological receptor are selected from the toxicity parameter data; Assessment factors are determined based on the type and quantity of the toxicity data; The soil ecological screening value is calculated by dividing the concentration of pollutants in the soil by the assessment factor.

5. The system according to claim 4, characterized in that, The determination of the evaluation factors includes: The assessment factor is 500 when there is at least one set of acute toxicity data L(E)C50 from plants, invertebrates or insects; When only single chronic toxicity data for plants or invertebrates are available, the assessment factor is 100. When there are two sets of chronic toxicity data that can represent four species, the assessment factor is 50; The assessment factor is 10 when there is chronic toxicity data that can represent at least three trophic levels and seven species.

6. The system according to claim 1, characterized in that, Based on the calculated comprehensive risk index Risk level assessment: : Negligible risk; Slight risk; Less risk; Significant risk; Serious risk.

7. The system according to claim 1, characterized in that, The interaction coefficients in the topology connection matrix Based on toxicological experimental data, it was determined that when antagonistic effects exist... The value ranges from 0.85 to 0.95; when a synergistic enhancement effect exists, The value ranges from 1.15 to 1.

25.

8. The system according to claim 1, characterized in that, The environmental impact coefficient Determined based on soil pH: When the soil pH is below 5.5, The value should be between 1.10 and 1.30; when the soil pH is between 6.5 and 7.5, The value is approximately 1.0; when the soil pH is higher than 8.0, The value ranges from 0.70 to 0.

90.

9. The system according to claim 1, characterized in that, The data management module includes a toxicity database, which contains toxicity parameters from domestic and international ecotoxicological data, relevant data released by national government departments, and reliable sources determined by experts. The toxicity database adopts the following data screening principles: Chronic toxicity parameters should be given priority, followed by acute toxicity parameters; When there are multiple toxicity parameters for the same toxicity endpoint of the same species, the geometric mean shall be taken; The formula for calculating the geometric mean is: ； in, for Each toxicity parameter value.

Images